You are on page 1of 124

Process Analysis

Via Accuracy Control


Maritime Administration
in cooperation with
Todd Pacific Shipyards Corporation
Form Approved
Report Documentation Page OMB No. 0704-0188

Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington
VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it
does not display a currently valid OMB control number.


AUG 1985 N/A -
The National Shipbuilding Research Program Process Analysis Via 5b. GRANT NUMBER
Accuracy Control





Naval Surface Warfare Center CD Code 2230 - Design Integration Tools
Building 192 Room 128 9500 MacArthur Bldg Bethesda, MD 20817-5700




Approved for public release, distribution unlimited



SAR 123
unclassified unclassified unclassified

Standard Form 298 (Rev. 8-98)

Prescribed by ANSI Std Z39-18
FOREWORD Since the original edition of this publication was issued in 1982, most U.S.
shipbuilders accepted the idea that appreciable productivity gains can be obtained by
more in-process accuracy and are responding accordingly. More importantly, some
also accepted statistical control of accuracy variations as the most effective technique
for control of work and for constantly improving productivity. They have responded
with significant investments, e.g., assigning college-educated people as Accuracy
Cmtrol (A/C) engineers and creating prerequisite databases.
In addition, the 1984 report, Toward More Productive Naval Shipbuilding:
issued by the Marine Board, National Research Council, related A/C to military
requirements, i.e., abilities to withstand high-impact shock and great submergence
To further support the A/C movement, the original issue of thk publication has
become the text for course-s in shipbuilding curriculums at the Universities of
Michigan and Washington. It serves further to indoctrinate both beginning and mid-
career people at the U.S. Navys Engineering Duty Officer School.
All of the foregoing created need for more in depth understanding of A/C which this
revision attempts to fulfill. Description of pertinent statistical theory has been made
more comprehensive. Use of the same principals which are the basis for Statistical
Quality Control, as advocated by Dr. W. Edwards Deming, is emphasized. A section
has been added on start-up which is based on actual experiences in U.S. shipyards.
Also, this edhion describes how a constantly improving manufacturing system
operates by providing an analytical basis without which Quality Circles are
ACKNOWLEDGEMENTS The material on which this publication was originally based was compiled by a
project team led by S. Nakanishi. International Division, Ishikawajima-Harima
Heavy Industries Co., Ltd. (II-II) of Japan. Team members included K. Ando and M.
Hatake. Their participation was managed by Y. Ichinose of IHI Marine Technology,
Inc., New York City.
The editor and contributing author of the original edition, February 1982, was L.. D.
Chirillo who assists the Los Angeles Division of Todd Pacific Shipyards Corporation
in management of research projects. He was assisted by R. D. ChirilIo and R. L.
Storch of L. D. Chirillo Associates and the University of Washington, respectively.
This revision was prepared by R. L. Storch with a number of passages added by
L. D. Chirillo- It is an end product of one of the many projects managed and cost
shared by Tdd for the Maritime Administration created National Shipbuilding
Research Program. The objective, described by Panel SP-2 of the Ship Production
Committee of the Society of Naval Architects and Marine Engineers, is to improve

Appendix D Samples of an Industrys and a Shipyards Accuracy Standards

Appendix E Analysis of Shrinkage in a Double-bottom Floor

Appendix F Control Charts


7his book is dedicated to the memory of
a shipbuilder
from New Orleans, Louisiana
Sigmund A. Sohzres
November 14,1933 May 31,1985

1.1 General

Effective shipbuilders accomplish in-house, only work which projects just as any other Iarge capital investment. Costs for
by nature and volume can be performed in accordance with the continuing the collection of data as a normal part of a production
concept of Group Technology (GT). Parts, subassemblies and process, are nominal because of the sampling techniques
assemblies, i.e., interim products, needed for an end product, employed.
are classified by the problems inherent in their manufacture.
Thus, regardless of differences in designs and quantities A/C cannot be effectively applied in the absence of aproduct-
required, interim products are manufactured on dedicated pro- oriented work breakdonm structure (PWBS) which fkatures
duction lines, i.e., process flows. A main flow. such as for interim products (i.e., fabricated parts and various subassem-
producing a hull block, is supported by coordinated subordinate blies) classified by the problem areas their manufacture
flows such as one for producing sub-blocks. imposes. This is the singular means used by the worlds most
effective shipbuilders to operate both real and virtual work flow
With such dependencies, a control system is required to lanes for a high variety of objects in mixed quantities.
monitor accuracy in order to avoid delays and rework, particu-
larly when blocks are joined together in a building dock during Because the different interim-products are classified by com-
hull erection. However, accuracy control (A/C) having a statisti- mon problem areas, the same work situations are sufficiently
cal basis does considerably more. A/C involves the regulation of repeated within each area for statistical treatment. Moreover, as
accuracy as a management technique for constandy improving sets of soIutions, e.g., specific classes of worker skih and
productivity of an entire manufacturing system. A/C is the facilities, we matched to problem areas, WC data are unaf-
scientific means which the worlds leading shipbuilders employ fected by variations that would otherwise occur.
for constantly identifying and prioritizing the problems which
must be addressed to obtain the greatest rates of productivity A final prerequisite for successful implementation of A/C for
improvement. The statistical basis makes clear the relationship hull construction, is the application of line heating for accu-
between cause and effect. As Dr. W. Edwards Deming teaches, rately curving and/or twisting plates and structural shapes such
the obligation to improve the system never ceases. as needed for regions of a hull that are curved. The need for
accuracy is criticaI in order to ebinate or at least minimize the
Statistics is the branch of mathematics dealing with collec- use of mechanical force when fitting components. When force is
tion, analysis, interpretation and presentation of masses of used, as with traditional shipfitting methods, structures develop
numerical data. The methods of statistics are methods of applied locked-in stresses and, following welding, distortion that is
mathematics. Shipbuilding engineers who manage A/C pro- neither predictable nor repeatable. A/C measurements and data
grams must at least understand college-level elementary statis- are meaningless in such circumstances and productivity is
tics. inherently limited regardless of degrees of experience possessed
by workers and their supervisors. Line heating by itself and also
Other prerequisites pertain to the data needed. An A/C data in conjunction with benders, presses and roIlers, permits curva-
base is a major investment. At first it requires systematic record- tures to be achieved with sufficient accuracy to eliminate or
ing of thousands of measurements. Such efforts are expensive. minimize force fitting.z
They will deter traditional managers having short-term goals.
These people are more likely to apply what they believe to be Thus, the major prerequisites for successful implementation
A/C as sporadic and unsophisticated preventive steps in of AIC for hull construction are:
response to one particular customers requirement for a specific
degree of accuracy. adoption of a product oriented work breakdown structure to
establish repeatable (standard) work processes regardless
Lack of long-term application negates the central importance of interim-product design differences,
of statistically-valid data which describes a shipyards normal
accuracy performances. Such data are the basis for continuing use of line heating for accurately shaping parts so as to
the collection of measurements by mathematically determined elimhate or minimize distortion after welding, and
sampling and for continued analysis and interpretation.
collection of an A/C data base describing a shipyards
Effective shipbuilders regard their A/C data base as a capital normal accuracy performances.
investment and means of production every bit as indispensable
as a crane or a building dock. The significant cost for starting an All three inextricably link accuracy and productivity.
AIC program makes sense only when it is amortized over future
1.See Product Work Breakdow Structure-Revised December 1982< National ShiphrildE Research program OWRP).
2. See Line Heating - November 1982T NSRP.

Some product-oriented shipbuilders evaluate each proposed
interim product or a lot consisting of more than one, for its
efficiency as a work package. Productivity Value (PV) is
expressed by the formula
PV = f(T,N,Q)
T= time allowed for its accomplishment,
N= number of units of resources, and
Q= quality of work environment and accuracy speci-
fied for the interim product.

The function (T,N,Q) is determined empirically and sepa-

rately for each stage within each flow lane. Each necessarily
considers the immedately preceding and following work stages.

Having PV vary directly with Q insofar as it applies to

accuracy specified for the intenm product seems to be a para-
dox. However, in this case Q relates to the efficiency of the
tolemnces specified with subsequent assembly work in mind.
Are the tolerances too accurate? Are they accurate enough? AIC
provides the method of determining the optimum tolerances
required at each stage consistent with the needs of customers,
regulatory societies and productivity.
When unprecedented projects are contemplated, A/C pro-
vides means for predicting how current work processes will
perform. When predcted productivity is not sufficient, man-
agers can determine the effects of changed design details, work FIGURE l-1: The A/C cycle is compatible with the management
sequences, and, if necessary, work processes, before work cycle of any industrial activity.
AIC is a repeating cycle of plan, execute, evaluate and replan;
Figure 1-L Vhal points and dimensions for blocks, sub-blocks Where most effatively applied, A/C engineers are assigned
and parts that are needed to assure accuracy of an end product throughout the operations department. Because their methods
are identified. They are systematically monitored at designated are analytical and always address the entire shipbuilding process
production stages. Similarly, many other measurements are their recommendations are inherently apolitical. Thus, they
made and carefully documented until scientifically valid samp- have the best opportunities for developing themselves as ship-
les of accuracy data have been collected. The data are evaluated buikiing engineers. As A/C experience is virtually prerequisite
using statistical methods to verify performance in terms of for higher managerial jobs, candidates are carefully selected
standard rcmges of accuracy normally encountered and toler- from people having about eight years shipbuilding experience
ance limits beyond which rework is required. By includlng such and memberships are rotated. This viable group, in addition to
written requirements in work instructions and by systematically its day-to-day planning, executing and evacuating, functions as a
monitoring, A/C tightens up all activities along a production defacto staff, i.e., advisory group, to the operations manager
line, e.g., template production, marking, cutting, bending, and his deputies even though they are assigned to different
fitting, welding, and line heating so that the tolerance require- departments at all managerial levels, including shops.
ments for each are compatible with the others. No longer are
crucial judgments about accuracy lefi to opinions and guesses. A/C provides scientifically derived, written and realistically
obtainable accuracy standards and goals. A/C is a function that
A specific example of tightening up for a particular work transcends departmental responsibilities. Whether it should be
process was further development of line heating to more accu- adopted should not be left to department or shop managers
rately form curved hull-parts as a means of rninitilng erection whose concerns are parochial.
wurk. Man-hours required for bending were reduced to almost
one tlird those needed for conventional rolling or pressing; A/C reports contain essential and reliable data that measure
fewer clips, dogs, wedges, etc. were required by assembly critical aspects of production performance and indicate where
workers; and rework for adjusting joint gaps during hull erec- improvements are required. Quite apart from controlling accu-
tion was greatly reduced. racy, A/C also defines management options regarding all
aspects of an operations organization. Implementation requires
total management commitment. In each shipyard, A/C should
significantly preoccupy the most senior operations manager.

FIGURE1-4: Example of an X and R Control Chart.

1.2,1. Control Charts

The most common tool used in statistical quality control is the

Shewbart control chart which was first described in 1924.
Control charts apply the concept of expected and measurable
variation in work processes, and are used to distinguish between
common and special causes. When only common cause varia-
tions are detected, the process is operating normally. Should
special cause variations be found, however, investigation and
identification of these causes is required in order to eliminate
them and return the process to normal performance.

measurements from a specific work process over time; see

Figure 14. Like standard deviation, the R for a random sample
is a measure of scatter. R is the difference between the largest
and smallest value in the sample. Although it is a less rigorous
measure of variability than the standard deviation, its simplicity
has led to its widespread use in control charts.
FIGURE 1-5: The Cemxat Limit Theorem.
Control chart theory is based on the statistical central limit
theorem. The theorem states that the distribution of the means of
random samples, taken from a normal distribution, is another
normal distribution with the same mean as the original distribu- Control charts are developed for a work process when the
tion, and a standard deviation equal to the standard deviation of process is in a state of statisticalcontrol. Some understanding of
the original distribution divided by the squareroot of the random the meaning of statistical control is important. A state of statisti-
sample size; see Figure 1-5. This result can be used to detect cal control is a state of randomness. When a process is in control
changes in the original distribution, which would indicate the
presence of special causes of variation.
charts will fall within the control limits. When Points fall outside
The technique involves initially determining regular perform- the control limits, they indicate the presenceof a special cause of
ance for a work process (i.e., its norrmd distribution), using a variation. The production worker can almost always detect and
large data sample. This normal performance can be used to correct these causes. When in a state of statistical control, a
establish an expected range of variation for the process. Subse- work process has prdctable and repeatable outputs. Thus, a
quent random samples of products from the work process can state of statistical control is evidenced by random sample values
then be monitored to detect changes in the performance of the
process. Control charts establish limits on the variation of the control limits and the sample size indicate the Ievel of accuracy
mean and range of these random samples. The liits are com- and the variation that can be expected.
monly set thee standard deviations above and below the process
mean and average range. Three standard deviation limits are Sound understanding of statistical control is essen-
used because they provide 99.7% assurance that exceeding these tial to management. . . . Stability, or the exist-
limits is the result of a change in the normal distribution of the ence of a system, is seldom a natural state. It is an
process and therefore the result of a special cause. Data such as achievement, the result of eliminating special
those shown in Figure 1-6 are used to determine individual causes one by one on statistical signal, leaving only
the random variation of a stable processes

The control charts provide information about a particular

work process. Since some variation is a regular result of any i.e., the centerline, upper control limit and lower control limit
work process, it is important to be able to distinguish between
expected change or random variations, and other variations. R charts are based on an established and repeated sampling
Thus the control charts are also used to assure that action need procedure. The sampling procedure includes a specified sampIe
not be taken to maintain the usually achieved accuracies of work size, n. The control chart values are determined from the results
processes. of a series of random samples.

W. Edwards Deming, "Quality, Productivity, and Competitive Position," MIT Center for Advanced Engineering Study, Cambridge, MA 1982, p. 119 (ISBN O-

1.2.2 Variation Merging

Ships are built by procuring or fabricating parts and then

joining them to create subassemblies. In turn, these are com-
bined through several manufacturing leveIs to produce increas-
ingly larger subassemblies, blocks and ultimately a complete
ship, Production line techniques are employed br the many
different interim products required.

When each of a succession of work processes is in statistical

control, its normal distribution of variations (mean and standard
deviation) can be determined. Whh such data, it is possible to
predict, statistically, the merged variation from the total series
of work processes.

Consider the combination of two work processes, cutting flat

bar and the spacing of longitudinal on a panel, Figures 1-2 and
1-3 respectively. The mean and standard deviation of variation in
fitting flat bars between longitudinal, Figure 1-7. can be pre- FIGURE 1-7 Effective shipbuilders add normal distributions of
dicted from the data of the individual work processes preceeding. variations from previous processes in order to predict how they
In addition to predicting the normal distribution of variation of merge at a later process.
the final process, the earlier work process which contributed the
most to the final or merged variation is identified. In this way,
knowledge of current work performance is applied by effective
shipbuilders to predict productivity for ship designs never
encountered before and, when necessary, to implement counter- processes necessay to insure needed accuracies of interim
measures before work starts. products, and to interim products themselves to insure required
accuracy of a final product, such as a ships hull. For the latter,
If there is need to reduce rework, accuracy goals are the merged variation, Z, is expressed as:
expressed in terms of the normal distribution required for the
final process. Then, by working backwards, necessary goals are
similarly set for each of the work processes which would insure
desired accuracy for the final process. Since normal perform- where:
ance at each work station is known, alternative kdding strate-
gies may be evaluated to determine if reduction in rework can be
obtained. If rework reductions by altering dasign details or
assembly sequences are not possible, steps to reduce the normal
variations for critical work processes can be initiated. These
may include improved tooling, better lighting, retraining work- This equation is referred to as the variation merging equation for
ers, or other such approaches. This product of A/C is called the completed hull.
process or method analysis. Process analysis involves a detailed
review of a particular work process. The goal is to reduce The variation merging equation is based on the theorem of
variability, i.e., shifting the mean variation and/or narrowing addition of variance. Variance is simply the square of the
the standard deviation of the variations of the process. A similar standard deviation. For independent distributions, such as those
approach can be applied to investigate special causes that are representing the normal performance of work processes, the
ponsible for a process being out of statistical control. theorem of addition of variance states:

Independent normal distributions, such as those representing

performance of each work process, can be added to determine
the expected normal performance at succeeding stages of con-
struction. Additions of normal distributions apply both to work the standard deviation of a final process.

1.2.3 Acceptance Sampling

In the same manner that effective shipbuilders apply accu- The choice of the level of acceptance sampling for incoming
racy standards to interim products manufactured in-house, they components is an economic one. The economic break-even
also apply similarstandards to materials and interim products in point for no inspection or 100% inspection is based on the
their vendors and subcontractors plants. Acceptance sampling
procedures are similar in theory to process monitoring using k1, and the cost to dismantle, repair, reassemble and test art
control charts. Suppliers control charts indicating their normal assembly that fails because a defective component was used, k z.
performances, permit inclusion of their data in shipyards perti- If;
nent variation merging equations even before purchase orders
are awarded. In other words, prudent shipbuilders require statis-
tical evidence of quality before issuing purchase orders for
materkds. When necessary, wise shipbuilders assist their sup- no inspection should be performed, and if:
pliers to implement and maintain A/C systems.
The use of special rolling facilities in some European ship-
yards is a perfect example of the need to integrate a shipyards 100% inspection should be performed. This results in a mini-
A/C system with a material suppIiers A/C system. Japanese mum cost approach to component acceptance sampliig.9
shipbuilders include steel mills statistical evidence of plate
flatness in their variation merging equations. As a consequence, Since this analytical approach is either not understood by
they have reached a state where developments in mills and politicians or inconsistent with their political objectives, gov-
shipyards have eliminated need for shipbuilders to have special ernment policies focus on identifying low bidders at the expense
rolliig faciMies. Shipbuilders who impose requirements for of obtaining assurances for quality before purchase commitm-
statistical evidence for the first time, will learn that some of their ents are made. Among the consequences are substandard
suppliers, steel mills in particular, already rely on statistical products and attendant cost increases. Great assurances and
control methods. lower total costs are obtained when a shipyard deals with fewer
suppliers, fostered to be proficient in MC matters and just
sufficient in number, e.g., three per material item, to insure
competition. Officials who make material procurement regula-
tions and shipbuilders themselves have to learn that shipbuilders
must deal with fever suppliers for productivity.~ reasons. The
alternative is continued disruption of shipbuilding efforts,

Standard ranges are indicated with the same plus and minus
notations used to fix tolerances. However, they are not toler-
ances. Instead, they are a measure of the actual accuracy capa-
bility of the processes used by a shipyard based on previously LOWER UPPER
collected data. A standard range reflects the accuracy range
currently obtainable with 95% probabihy, for a particular work
process. Tolerance limits required for interim products and end
products, whether for productivity or quaIity/functional consid-
erations, should encompass the associated standard range, as
shown in Figure 2-1. Where they do not, rework can be regularly

The use of ranges and limits as described in the foregoing is

proven and acceptable to classification societies. Such use and
continuing analyses of data enable Japanese managers to know
where they are regarding accuracy being achieved and where
they stand regarding acceptance. They know what they have to
do next to improve their shipbuilding methods. Their abilities to
regulate accuracy are a powerful means for managing shipbuild-
ing operations.

The disciplines for producing statistical evidence are particu-

larly credited by Japanese shipbuilders for tremendous improve-
ments in productivity. In 1967 they reported in English that
statistical control of manufacturing epoch makingly
improved the quality of hull construction, laid the foundation of
modem ship-construction methods and made it possible to
extensively develop automated and specialized welding.z
Statistical analysis of accuracy variations of a shipyards cur-
rent work processes can be used to predict how accurate hull
structure will be in a ship never built before. This has great
significance for owners, particularly the Navy. Abilities to
withstand high-impact shock are directly related to accuracies
achieved without forced fitting during construction processes.
Maximum submergence depth of a submarine is related to the
degree of hull circularity achieved and absence of locked in
stresses. Thus, the Navys possession of statistical evidence of
accuracy from shipyards before award of contracts, would serve
military requirements?
As quality and productivity are directly related and since A/C
provides an analytical basis for less direct inspection, there are
prospects for savings by both shipbuilders and owners. Owners,
shipbuilders, and suppliers need to fimther exploit statistical
control of manufacturing.

An important aspect of A/C is the difficulty commonly

encountered in joining blocks during hull erection. Erection
joint gaps that are not within tolerance limits must be reworked
by gas cutting and/or buildup by back-strip welding as shown in
Figure 2-2. Effective shipbuilders have proven that applying
A/C to all earlier work processes is more productive than having
to deal with excessive merged variation in relatively inaccessi-
ble and hazardous locations at a building berth.
Everyone involved in implementing A/C understand that Initially estimated tolerance limits and excess allowances
only when work is in control, can statistical distributions of are included on work instructions. With more data and analy-
variations be used to predict with confidence the quality and sis, the limits and allowances are refined and revised. As the
productivity for a contemplated end product having design de- data base grows it will be more and more relied upon as a man-
tails never encountered before. Sophisticated customers require agement tool and its status as capital become indisputable.
such statistical evidence of quality and productivity before con-
tract award. As such predictions direct attention to specific work Written procedures for work processes are essential prior to
methods which when improved do the most for improving the any data measuring. Following the same procedure per work
productivity of an overall manufacturing system, A/C is the sin- stage is essential in order to achieve a normal distribution of
gular element of competition among the most effective ship- variations. Particularly in product organizations where speciali-
builders. Failure of a traditional shop manager to cooperate will zation is by problem categories, developing work procedures
have serious consequences because of the increasing complexity causes an interaction between production engineers, designers
of shipyard end products and because of shipbuilders urgent and shop people during which each learns more of the others
needs to become more flexible in market places. In the most requirements. As emphasis is placed on constant improvement
effective shipyards, understanding statistical analysis and con- in methods, these work procedures are standards of the mo-
tributing to A/C systems are prerequisites for shop management. ment which are revised as soon as better methods are manifest.
Thus, the vital interaction of production engineers, designers
During such indoctrination, emphasis should be made of the and shop people prodded by continuing directions emanating
tentative nature of tolerance limits per stage because when work from statistical analysis, continues forever.
becomes in control, consistent with statistical theory the limits
are regarded as three standard deviations. That is, when work is Data collection should be a production responsibility
performing normally, probably 99.7% of the items being worked equivalent to marking, cutting, fitting, welding and distortion
will be within tolerance limits and probably only 0.3% will re- removal. A typical approach might involve unrecorded meas-
quire rework. When performance of work changes so as to af- urements of all products by each worker followed by recorded
fect a normal distribution, the tolerance limits change accord- measurements based on random sampling by a first-line super-
ingly. visor and again by a second-line supervisor. What is most
amenable for each shipyard should be adopted, as long as
If tolerance limits are not acceptable for customer specified measuring is a shop responsibility. As for any other shop work,
requirements or for facilitating work downstream, such as for allowances should be made in schedules for measurement
welding of erection butts and seams, management responsibility work.
is automatically acknowledged and resolution may involve de-
sign modifications, changes in work sequences, additional Measurements serve multiple functions. They facilitate
training for workers and/or inventing a new work process until achieving the short-term goal of minimum rework during hull
tolerance limits are acceptable. In other words, by definition and erection. Thus, during start-up, one set of measurements
with 99.7% probability, no requirement is imposed that workers should follow an interim product through its work stages.
cannot fulfill through their normal performance of work. Measurements and tolerance limits preclude arbitrary accumu-
lation of variations since unacceptable work is not passed on to
Both the iterative nature of and time span required to achieve later work stages. The data also provide statistical performance
a working A/C system must be made clear to everyone involved. indicators for incorporation in the A/C management scheme,
Implementation throughout a shipyard is difficult but necessary. i.e., an A/C analysis group determines the mean and standard
A choice of one or two specific areas will ease start-up. Appli- deviation per work stage. An extremely important part of an
cation to parts and assemblies for simple structures, such as, for A/C system is to publish the analyzed data in a form that eve-
double bottoms and wing tanks, is an ideal early choice. Later ryone, workers included, will readily understand. This func-
more complicated structures and outfitting work, including that tion, essentially feedback, holds the key to achieving positive
assigned to subcontractors, should be addressed based on results returns from investment in an A/C system.
of initial attempts. Responsibilities for specific tasks within an
overall framework must be assigned. Subgroups, made up of Traditional production people can no longer react to prob-
people from production engineering, design and production lems without regard for the informational requirements of an
shops should evolve naturally if a product organization is in A/C system. The best solution to a problem encountered might
place. These groups will develop check sheets and initial se- be in design, material procurement or in an earlier work stage.
quences, tolerance limits and excess allowances. As such infor- Thus changes and problems must be documented and resolved
mation is developed, individuals having primary responsibilities, with production engineers and designers as dictated by statisti-
such as loftsmen, gain insight and are able to continue with cal analysis. Unless benefits are overwhelmingly manifest,
fewer and shorter instructional meetings. changes in work procedures per work stage should only be ac-
complished when the impact on the entire manufacturing sys-
tem is determined by statistical analysis. 4

Concerning descriptions in writing, a Sicilian proverb says, White soil, black seed. Beware of the man who sows it. He never forgets. And so it is with A/C because
real performance data are recorded and continuously analyzed in behalf of an entire manufacturing system. It constantly targets problems, it monitors rates of im-
provement and it never forgets. Traditional shop managers are apt to be apprehensive. Ineffective shop managers have reason to be apprehensive.


A/C execution is concerned with: defining who measures, work loads are 1also posted. Self checks, subsequent cheeks
when and how measurements are made, and recording data. and recordings are regarded as work processes that must
adhere to these schedules . A blackboard in each division of
production shows the day-today status.
4.1 Self Check
Normally, the master schedule for block ereeticnq weekly
A/C includes a self-check system which workers and their progress sheets and a schedule for erection cheeks based on
immediate leaders exeeute. Self checks are erueial. Workem the master schedude are posted in an erection office. The day-
have not completed a job until they have checked their work today status of block erection is maintained on a black-
to assure compliance with written accuracy instructions. board.
Thus, self checks are regarded as work just as much as any
other work task. Subsequently, work leaders, one for approx- Accuracy checks are performed daily in accordanee with
imately every eight workers, cheek the same work andrecord schedules that am revised weekly, if necessary. Basically, the
the pertinent final data accordingly. Very important check items checked for conformance with accuracy standards are
points and lines, i.e., control items, are again cheeked and
recorded by the next higher level of supervision. If such data
are unreliable or not available there is no point in having WC. exeess allowances and marks required for fabrication,
assembly and checking work,

4.2 A/C Group

shapes, edge preparations, deformation, and the cur-
Where WC is successfully applied, people having responsi- vature of bent parts,
bilifities to execute A/C procedures are assigned in the hull
construction department. All are members of a yard-wide
A/C group, have 8 to 9 years of varied shipbuilding experi- parts or sub-blocks, their fit, gaps for welding distortion
ences and were carefully selected on the basis of their aptitude and overall dimensions, and
for and commitment to improving productivity. Their
responsibilities are
hull alignment.

be just dependent on the self-check system. 4.4 Information for Check Sheets
In accordance with work instructions issued by designers
and based on information provided by A/C planners, mem-
bers of an A/C group in a hull-construction department pre-
pare check sheets. These designate check points and lines,
checking methods, responsible personnel for meaning, and
required frequency for measuring. Typical examples of cheek
sheets are incorporated in Appendix A.
Preparing check sheets for curved blocksis usually difficult
because the dimensions included in normal working draw-
senior operations manager and attended by the managers ings, while sufficient for assembly work, are not suitable for
and deputies of the major ditviilons of the operations checking purposes. The simplest example are the two
department, for discussion of v0roductivity matters. diagonals required for verifying the mtmguhity of a panel.
The A/C group advises loftsmen to calculate numerous other
As participation in rovides an excellent overview of special dimensions that facilitate accuracy checks; examples
planning, executing and evaluating, A/C group experience is of these are shown in Attachments 4,5 and 6 of Appen-
prerequisite for higher managerial responsibilities. And, dix A.
because increased produetivig is dependent on more
managers acquiring a complete overview of the entire ship- Actual measurements are mainly performed as specified by
building proms, memberships in an A/C group are rotated. the check sheets. However, check sheets cannot practically
provide for all dimensions for all hull parts and assemblies.
There has to be some dependence on supplementary routine
4.3 When and What to Check checking of other dimensions by workers. This helps insure
that the dimensions required by cheek sheets will satisfy
Usually, schedules are posted for starting and finishing accuracy standards. Typically, check sheets address dimen-
dates at each control station for part fabrication, sub-block sions and measuring methods as briefly illustrated in Figures
assembly and blink assembly. Summary sheets for future0 4-1 and 4-2.

FIGURE4-1: Dimensions and check methods that are typically W subject of check-sheet instructions for upper-wing tank parts in parallel
midbody. See legend in Figure 4-2.

4.4.1 Part Fabrication

In order to achieve specified accuracy during assembly

work, each of many parts must be fabricated within specified
accuracy standards. As measuring every dimension of every
part is impractical, random sampling is employed to monitor
accuracy tendencies. However, special or large structural
parts, such as girder or transverse web assemblies are excep-
tions. Each should be measured meticulously per check sheet
instructions with particular attention to deformation. When
cutting machines, such as N/C, are employed, their mainte-
nance is a significant factor in the uniform working circum-
stances which are the bases for a valid random sampling.
Maintenance checks on cutting machines should be frequent
and regular.

The accuracy of bent parts is critical for achieving the ac-

curacies specified for assemblies. Inaccurately bent parts are
frequently forced to fit and are the sources of internal stresses
which cause deformation when welding. Thus, all crowed
shell parts should be checked using sight-line templates and
other information provided by loftsmen in order to establish
for each plate as required

FIGURE 4-3: As shown, each template is set at a specified station

and angle with its edge marks matching plate edges- Sight line
marks, relative to a taut string, determine the accuracy of
templates, longitudinal curvature. Inclination of the sight line is an aspect of
human engineering required as an A/C measure. A checker main-
tains an efficient, relaxed stance. Such techniques when repeated
many times, significantly contribute to increased productivity.

Regardless of their shapes, blocks are categorized by the

panel upon which they are assembled, i.e., flat or curved.
Analogous techniques and checks apply to other parts such as Typically the former are assembled on flat datens and the lat-
twisted Iongitudinals. ter on pin jigs. Measurement methods for the the types are
..- necessarily different.
4.4.2 Sub-block Assembly
Flat-block check sheets should include the following
Typically, what is important for A/C of sub-blocks is the requirements:
fit of stiffeners, brackets and face plates such as on a web
plate, and how to prevent and/or deal with deformation and
shrinkage caused by welding. Therefore, measuring activity just after the base panel is assembled,
during sub-block assembly should concentrate on:

checking fitting dimensions,

checking for deformation and shrinkage by using a base panel is completed, and
reference line on a web plate and/or a straight edge of
the web plate, and
unique aspects of flat blocks which incorporate some
measuring other dimensions as indicated on a check curved shell.
4.4.3 Block Assembly

Achieving specified accuracy in an assembled block is most

important because the block assembly process offers the last
opportunity to deal with variations that otherwise have to be
considered during erection.

caused by welding, and

panel edges and the edges of internal structure particu-

larly near erection joints.

Checking blocks as described in the foregoing is important

because many are neat cut along erection joints during the
/4 * -/ final phase of block assembly.
FIGURE 4-4: A and H are typical vital dimensions which A/C
engineers require. Usually they are provided for in work instruc-
tions prepared by designers; Ioftsmen calculate their actual values. 4.4.4 Hull Erection

Curved-block check sheets should include the following During the erection stage, the object is to at least achieve
requirements: end-product accuracy standards specified by regulatory
societies and owners for hull depth, breadth, length and
measurements to check guides for precisely locating curv- straightness. WC group members- monitor vital points and
ed plates for a base panel on a pin jig, dimensions by measuring and recording periodically per
check sheet instructions during the entire period between keel
measurements of width, length, diagonals and chord laying and Iauncm see Figure-4-5 and Appendix A.
lengths to be made just after a base panel is assembled,

use of marked steel-tapes prepared by Ioftsmen for

checking assembly finished-marking, i.e., the location s
of sub-blocks and internal parts on a curved panel,

BOllOM Nov. 17
Time 930 am
Temp. 16C


F10 F39 F45 F54 * F62 F69 F75 F81

203 mm below
< bottom line

( ~ ) (y? ( q (::) q )

FIGURE 4-5: In order to achieve end-prcxluct accurancy, WC engineers carefully monitor the alignment of assembled blocks throughout
tbe entire hull erection period. Usually, regulatory and/or owner representatives witness these activities.

Analysis is the foundation upon which A/C is built. How to 5.1 Regular Analysis
analyze is the most important prerequisite for shipyard manage-
ment. Once an A/C system is in place and functioning, monitoring
work performance per processor stage simply involves random
Systemized A/C analysis and feedback ensures that experi- sampling and plotting of results on control charts. Workers and
ences and lessons learned are acquired by the organization and first-line supervisors monitor the charts to watch for changes in
translated into improved productivity. As work progresses, all work performances. Analysis of control charts beyond identifi-
results from check sheets and reported accuracy problems are cation of special causes of variation is adequately covered in
analzed by the A/C group before they are sent to concerned
organizational divisions. The evaluations include
If an analysis discloses an apparent area for improvement an
analysis, and A/C engineer pursues one or more typical options as foIIows:

recommendations which, as shown in Figure 5-1, are per- more detailed investigation of the data,
formed on either a regular or an urgent basis.
investigation of instruments used for measuring,
Regular analysis is employed during a number of phases, even
during start-up. Typical regular analysis functions include verification of alignment of facilities such as platens for
flat-block assembly and cribbing for erection,
. determining normal performance by process flow or work
stage during start-up and following changes in work meth- review of review methods, and
study of specified amounts for excess.

stage, Workers perform self checks daily to insure compliance with

accuracy standards. These are again checked and recorded by
their work leaders. Properly collected data, even if all measure-
stage, using a pre-established sampling plan, ments are within accuracy standards, are used to identify the
characteristics and tendencies of variations. Such knowledge
leads to further improvement in production processes. An
on design details, work sequences, a block plan, etc., and example of data collection and analysis for determining excess
allowances is included in Appendix E.

variation merging equations, for identifying specific proc- Feedback of analyzed A/C data is vital because it encourages
ess flows or work stages which if modified would fbrther planners to review matters such as:
improve productiviq of the entire manufacturing system.

Urgent analysis takes place when sampling indicates that an dimensions. etc., were satisfactory,
interim product is not width tolerance limits and therefore has
the potential to disrupt ensuing work. Urgent analysis is used to
quickly determine the best remedial action, e.g., immediate
rework and rescheduling of succeeding work within limits
imposed by an erection schedule, providing compensation by
changes in design details or work processes for other interim
products, and/or initiation of overtime work.

See, for example, Introduction to Statistical Quality Control; by D. C. Montgomery, Johrr Wiiey & Scars, 1985.
5.1.1. Significance of Mean Vhhte
For most work processes, the mean value for variations is Data analysis quantitatively sets amracy standards. Rx
planned to be zero. If the actual mean value differs from zero, it example, when erection joints are aligned the achieved dismti-
should be changed to match results of the work process or the tion of gap variations will, at the extremities of the distribution,
work process shouId be changed so as to yield the planned mean show requirements for rework
value (zero). The following examples apply:
. cutting where a gap is too small or negative, or

such as for a longitudinal bulkhead under a tank top, which . building on an edge where them is too much gap.
were cut with some allowance for shrinkage. After welding
during sub-assembly work, the mean value of the dimen- As shown in Figute 5-2, when G, is less than O, minimal
sion was determined to be negative, i.e., some shortage material is cutoff to achieve the gap G. because it is cheaWr to
exists compared to the planned zero value. retain as much of the original material as possible. When G. is
more than O, a minimal amount is built-up to achieve the gap G W
Analysis: Check kerf compensation; if sufficient, the because the buildup proms is expensive. Thus, G. is always
allowance for shrinkage was too small. smaller than G..

Remedy: Add the absolute mean value to the previously

planned allowance for shrinkage. Therefore, by definition the lower tolerance hit is G, and the
upper tolerance limit is G.. A standard range to be used as a goal
Example 2: Near the end of flat-block assembly, checking for improving Gaeanbeestablished accordingly; see Figure 5-3.
discloses that plates in tank-top panels are deforrmd at their
centers with a mean value of 1/2 inch. em Buildup

Analysis: Check the level of the platen on which the flat

blocks were assembled.

Remedy: If the platen is true, improve the assembly work

processes, e.g., apply pre-tensioning or change weld a.sck Ship
sequenees. 4

5.1.2. Significance of Standard Deviation Ga - initial

Gn - gap after rework by gas cutting
Standard deviation is significant for a number of reasons. It Gw - gap after rework by back-strip welding
provides linkage between the accuracies of earlier work proc-
esses and the accuracy of a final process through the theorem of FIGURE 5-2: AIC is most effective when it ftxuses on minimiz-
addition of variance. Without this relationship, analytical A/C ing the two kinds of rework commonly enmuntered when joining
does not exist. hull blocks, i.e., gas cutting and buildup tg back-strip weMin

Further, during analysis A/C engineers are very watchful for a

change or shift in the standard deviation for each work process. Gn Gw
Such behavior could indicate that something about how a work
process is executed has changed. Many reasons exist including a BACK-3TRIP WELO
worker perfecting abetter technique which should be adopted by
others and erratic operation of or deteriorating machinery.
. Example: The standard deviation for the length of manuallyl
fabricated Iongitudinals suddenly increases, decreases or

Analysis: Examine how and by whom the longitudinal

were fabricated. Methods, particularly sequences, should
be thoroughly analyzed.

Remedy: There could be many solutions dependent upon Gn - lower tolemnee limit
results of the detailed analysis. At least one shipbuilder Gw - upper tolerance Iimlt
reacted by finish cutting Iongitudinals before bending, i.e.,
end-margins to permit grasping for bending at the ends,
were eliminated. Following the mechanical bending proc- FIGURE 5-3: The lower tolerance limit stems from having to
ess, line heating was introduced to bend the finish-cut ends. create an erection-joint gap or make it wider. The upper tolerance
Accuracy was improved and the wasteful margins were limit arises from having to make a gap narrower.
5.I.4. Modifying Distribudons

Consider traditional rework for adjusting erection gaps. Cut-

ting dominates because experienced people know that cutting
costs per lineal measure are less than costs for build-up by back-
strip welding per lineal measure. The mean value of the perti-
nent distribution of gap variations favors the lower tolerance
limit accordingly. Figure 5-4 shows this intentional bias and also
shows the impact of shifting the mean value toward the upper
tolerance limit.

Because of the nature of normal distributions, the nominal

increase in back-strip welding is overwhelmingly offset by the
substantial decrease in cutting required. Further, the prospects
for exceedng the lower tolerance limit are reduced. Thus,
analytically derived goals proposed by A/C engineers will
sometimes differ from those adhered to by traditionalists who
operate without the benefit of carefully collected and analyzed

In shipyards where A/C is practiced, operations managers

benefit from detailed reports of productivity during hull erection
which relate total lengths of gas cutting and back-strip welding
to the total lengths of erection gaps. In an actual report for
erection of a 167,500 DWT tanker, Figure 5-5, rework was only
required for 32.6% of total gap lengths. That is, 67.4% which
had been neat cut was already within tolerances for semia-
automatic or automatic erection welding.

A similar report for a 32,000 DWT bulk carrier, Figure 5-6,

shows that rework was required for 38.7% of total gap lengths.
This higher rework figure is due to the smaller ship having
relatively more shape. The actual gap widths achieved and the
locations of rework by gas cutting and buildup by back-strip
weldlng are shown in Figure 5-7.



l! I
/ 1
x x,

5.1.5 Sequence for Analysis Reportedly, the average time for such meetings is short; for the
most extreme problem two hours coald be required.
Ongoing review of accuracy standards by continuously ana-
lyzing data is very important. The following procedure for After the temporary countermeasures for quickly restoring
analysis of data obtained during flat-block assembly is typical: work flow, investigations continue for the purpose of devising
permanent solutions. Usually, work procedures are revised to
reflect more A/C philosophy.
teristic, e.g., length, width, etc., as shown in Figure 5-8,

each characteristic,

conforms to its pertinent standard range, e.g., competitive

standard range (X+2S means conformance with 95 % prob-

ability), A/C engineers:

- confirm that the standard range is appropriate, investi-

gate and make necessary recommendations, e.g., adjust
excess allowance, change methods, supplement worker
. training, etc., or

- propose changes in the standard range which do not

impact on end-product tolerances.

Appendix E contains a good example of a sequence for analysis.

5.2 Urgent Analysis

In real shipbuilding circumstances no one can eliminate varia-

tions which require rework. Moreover, no one can predict
exactly when they will occur. Disruption is also caused by the
effects of such things as errors, accidents and weather abnor-
malities, which differ from variations because their occurrences
do not adhere to normal distributions. Despite their erratic
natures they too require organized responses and analyses in
order to:

disruptions, and to subsequently,

The feedback path for these urgent considerations is included in

Figure 5-1.

One shipbuilders preplanned response to a serious inaccu-

racy immediately summons select members of the A/C group.
This trouble-shooting team of specialists for planning, execut-
ing and evaluating, meet where the inaccuracy exists to:
evaluate impact on work flow,

recommend what, how, where and when rework is to take

place so as to minimize disruption, and

collect evidence for identifying the cause.


$ 20-
0 N= 146
z Width x = -0.33
S = 1.66
$ lo- S/ * 2s:-4.05

- 9 - 6 - 7 - 6 - 5 4 -3 -2-1 0 1 2 3 4 5

Squareness N= 1 5 0
X = 1.61
s = 1.09
E + 2s:3.79

0 1 2 3 4 5 6 7 6 9 1 0 1 1 1 2 1 3 1 4 1 5

N= 150
z = 4.43
s = 3.05
x + 2s:10.53

FIGURE5-8: Each histogram of variations in flat blocks compares identical parameters. Regarding squareness and flatness only absolute
values are of concern. Append~ B contains more information about preparing histograms.

Measurement Sample Standard
Frequency Deviation
Stage Regular Control Rem
Week 20 0.4
Template Tape
ProductIon 20 Days 8 0.5
Paper Template
20 Days 8 0.4
Tin Template
20 Days 8 0.5
Wood Template
Cuttin Mate by Flame 0.4
Part planer-Width Day 8
Cutting Plate by Flame
planer-Straightness Day All
Finish Marking plate
Day All
Finish Marking Plate
Main Marking Line Day All
Finish Marking Plate
Day All
Right Angle
10 Days 8 1.0
Bevel Angle for Auto Welder
Day 8 0.8
curved Plate Marking
Day 8 0.8
Cutting Accuracy of Curved Plate
Day 8 0.5
Shape Marking
Day 8 0.8
Cutting Accuracy of Shapes
N/C Cutting Machine 0.6
Day 8
Plate Width
N/C Cutting Machine 0.5
Day 8
Plate Length
Cutting Accuracy of Internals 8 0.8
e.g., Floor Girder in a Double Bottom Day
Day 8 1.5
Cutting Accuracy
Day 8 0.7
Sub-Block Accuracy of Fitting Stiffner
Assembly Straightening Deformation
Day 6 0.8
by Line Heating
Day 8 0.8
Accuracy of Fitting Face Plate
Day l/200
Accuracy of Fitting Angle
2 Days 8 1.4
Block Plate Length
2 Days 8 1.5
Assembly Plate Width
Right Angle (Difference between 8 12
Diagonals) 2 Days
2 Days 8 0.8
Reference Line
2 Days 8 1.2
Position of Longitudinal Edge
2 Days 8 1.5
Position of Transverse End
2 Days 6 1.5
Accuracy of Through Piece
2 Days 6 5/1000
Accuracy of Curved Shell Web
Curved Shell Plate-Length 1.5
3 Days 4
(After Cutting)
curved shell Plate-Width 1.5
3 Days 4
(After Cutting)
Curved Shell Plate-Reference line 8 0.8
(After Cutting) 3 Days

FIGURE 5-9: Regular Control Items.


5.3 Control

Controls which assure that achieved accuracy conforms with

an A/C plan for hull construction, is prerequisite for competitive
shipbuilding. They are classified as regular or special.

Because of the many different parts and subassemblies that are

required, regular controls are applied to repetitive mork pro-

gram, including their measurement frequencies, sample sizes

and standard deviations, are listed in Figure 5-9. A control chart
for such regular usage is shown in Figure 5-10. Such charts are
maintained by A/C engineers for production control purposes.
Once people become used to them, they provide guidance to
everyone concerned, i.e., workers and their supervisors. Thus,
each such control chart is posted at its respective work station.
This is important. Descriptions of the types of control charts
used for AIC by shipbuilders and how to prepare them, are in
Appendix F.

Special controls are based upon the accuracy condition of a

hull upon completion. Necessary vital points are defined and
included in the A/C plan for a specific hull. When the hull is
completed, members of the A/C group accumulate and analyze
measurements that relate to predetermined vital dimensions.
They look for accuracy trends which should be modified for
further productivity improvements.
Use of variation merging equations permits prediction of
probable erection gap accuracy as described in Figure 3-4
However, more information is needed in order to predict the
locations of and amounts by which gaps are out of tolerance for
semiautomatic or automatic erection welding. With such predic-
tions, rework is planned accordingly. The decisions made
include where and when rework shall take place. e.g., gas
cutting could be performed by either assembly shop or erection
shop workers dependent on circumstances.

Measurements needed for rework planning are made after

block assembly to determine how block widths and lengths vary
from design dimensions. Organizing such as-built measure-
ments, as shown in Figures 5-11and 5-12, is sufficient input for
some very experienced A/C engineers to make rework deci-
sions. Others find it helpful to employ position Dimension
Diagrams (P/Ds) for checking gap accuracy achieved against
accuracy needed to facilitate erection welding.


A z

closer together.

5.4 Process Analysis

Effective control of accuracy is dependent on proper under- Since these equations indicate the contribution of individual
standing of variation merging equations such as those given in work process variability to final or-merged variability, prime
Figure 3-4. Too much focus on a merged variation, Z, is not candidates for process or methods improvement are analytically
worthwhile. It is more important to focus on each fictor on the pinpointed. Such methods of analysis, to reduce process varia-
right side of each eauation. If these factors are sufficiently bility, can involve consideration of new tools. additional train-
controlled, nominal checks will suffice to confirm each merged ing,- modified work environment, etc. Employing statistical
variation. Some of these nominal checks, usually by sampling, techniques, effective shipbuilders witnessed the advent of spon-
are useful for balancing alternatives such as gas cutting vs. taneous quality circles which they subsequently trained to apply
back-strip welding as shown in Figures 5-2 through 5-4. formal analysis techniques. Such efforts lead to constant break-
throughs in productivity improvement, i.e., for a work process,
The alteration of a design dimension to control the amount change in the mean variation and/or shift in the standard devia-
and type of rework is a form of process analysis. tion of variation as shown in Figure 5-15.

Based on the variation merging equations, additional types of

process analyses are possible. The impact of altering assembly
sequences can be evaluated in terms of rework types at succeed-

For a detailed example of this type ofadysis see Accurscy Control Variation-Merging Equations: A Case StUdy of Their Application in U.S. Shipyards: Richard
Lee Storch, The Journal of Ship Production, The Society of Naval Architects and Marine Engineers, May 1985 (ISSN 87564417).


6. 1 Design
The hull-block construction method developed naturally f&- Further, qualified people should be assigned as specialists to
lowing the introduction of welding many years ago. Some do the checking. Loft defect lists and graphic representations
shipbuilders changed their organization of structural drawings of frequency of occurrence, as shown in Figure 6-1, are con-
trol mechanisms used by A/C engineers.
tion plan, block assembly plan, sub-block assembly plan and
Each mold loft should be regarded as a nucleus for A/C
drawings because they associate classifications of parts and activities because it generates most of what is used both for
assemblies with specific manufacturing levels in production. achieving and maintaining a specified degree of accuracy.
They are, literally, work-instructions. Loft processes for producing N/C data, templates and other
information formats shouId include essential A/C require-
Design and material definition should be truly regarded as ments such as:
aspects of planning and drawings should be further developed as
virtually complete work instructions including A/C work. When
A/C requirements, particularly vital points and dimensions and
excess allowances, are included:
calculated special dimensions which facilitate assembly
checking and recording are clearly delineated as work just and checking work,
as much as marking, cutting, fitting, etc.
reference lines and check points,
excesses to compensate for shrinkages are adequately con-
sidered and are consistently applied, and adequate marks for lay out marking (while most are suf-
ficient for snapping a chalk line, there is difficulty in
the potential for human error is reduced; loft, fabrication identifying which marks associate with each other).
and-assembly workers no longer have to refer to separately
prepared A/C requirements or depend upon recollections. excess already incorporated (e.g., when workers do not
have to separately mark an excess allowance, A/C is
outputs, designers should also be required to respond to A/C
analyses. A/C feedback includes variation reports, design and more sufficient bridging instructions to minimize warp-
process improvement recommendations and updates to the A/C age and shrinkage during gas cutting.
data base.Designers should becollateral responsibili-
ties to participate in A/C analyses particularly as they apply to Production Control
evaluating the affect of design alternatives on productivity.
Further, designers should be alert to promptly update detail If just the terms part fabrication, sub-block assembly and
design standards commensurate with a shipyards changing block assembly are coded in a marking system for interim
accuracy capabilities. products, a relatively modem innovation to some it is diffi-
cult to relate an explosion of vital points to an explosion of a
6.2 Mold Loft hull into interim products. Further classifications of such pro-
ducts should be included in a marking system so that each in-
Strictly speaking, loft processes should be subject to the terim product has a unique identity, e.g., by zone. In other
same A/C scrutiny as marking and cutting in a part fabrica- words, a fully deveIoped product-oriented work breakdown
tion shop. However, mold-loft process variations are too structure is essential for effective A/C planning, executing
small to significantly impact on merged variation during part and evaluating.
fabrication. But, loft errors (mistakes, omissions, etc.) are of
concern because they disrupt the A/C cycle. Via product orientation, designers can respond more read-
ily to production control requirements for work instructions.
Errors cannot be treated with classical A/C theory, i.e., The latter are more than just detail drawings because they
they do not enter into variation merging equations. There- define interim products and specific sequences for their
fore, for A/C purposes written procedures shouId be devel- manufacture. With information so organized designers can
oped in order to address: more readily respond to A/C requirements to include, for
example, tolerance limits and vital points in work instruc-
classifications of errors, and tions. Providing such information in work instructions,
because they are the most universality empIoyed documents,
methods for checking, recording and analyzing (the sta- facilitates mutual understanding of A/C requirements and
tistical principles described in Appendix B could be more efficient execution by loft, fabrication and assembly
used). workers as well as by members of the A/C group.

In addition product orientation permits sufficient classifi-
cation of the myriad of part and subassembly geometries in
order to relate them to specific work prcnsses. This associa-
tion is critical for obtaining valid A/C data. Otherwise, work
circumstams are insufficiently controlled and virtually no
data sample will approximate a normal distribution. A/C as a
science would not be applicable.

6.4 Fabrication
Measured after gas
N/C gas cutting is almost universally applied by shipbuild- cutting
ers but there are still situations where semiautomatic cutters
are useful supplements to N/C installations. More variation is
probable in a semiautomatic process, therefore, A/C require-
ments should be different. However, there are common
considerations when accuracy performances need to be

advanced N/C systems,

Measured after
plate welding
different part classifications, e.g., parallekdge part, in-
ternal part, etc.,
l kerf tolerances should be specified,
FZGURE 6-2: Work sequences impact on total shrinkage. If the
an N/C machine, should be performed regularly and fre- two plates m the lower figure were welded together&fore sss cut-
quently, worn torch-tips should be replaced and others ting, the !cmver figure would reflect only shrinkage due to gas cut-
cleaned, ting.

solved, even where shipbuilding technology is most ad-

vanced, measurement data should be accumulated on the
effect of different cutting sequences, bridge restraints,
etc., and

e.g., lasers which can be focused, could perform with

narrower kerfs, less heat input artd thus less shrinkage
and distortion.

Further, deformation such as caused by welding should be

6.5 Sub-block Asembly diminished by pretensionirtg and/or removed by line heating.
Methods to control deformation, such as pretensioning, Fitting processes for sub-block assembly are mainly per-
preheating and spedfkd welding sequences, should be prac- formed manually. Where A/C is ongoing, there is irtdispens-
ticed. Regarding shrinkage, consider the panel for the sub- able close association between supervision of sub-block
block shown in Figure 6-2. When the large plate is gas cut, assembly work and the A/C engineer assigned to the sub-
shrinkage Al, occurs because no bridges were provided across block assembly section (perhaps in a small shipyard assigned
the cutouts. Additional skrinkage A2 occurs when the large to the hull ccmstruction department). Because of preoccupa-
plate is welded to the small plate. Without a shrinkage allow- tion with variations in each work process and how they
ance, the combined shrinkage Al + A2 could necessitate merge, the A/C engineer readily detects situations where simp-
rework, i.e., making the cutouts deeper during block assem- le jigs enhance both accuracy and productivity. More such
bly. jigs should be used.

6.9.3 Ways to Distribute Exess vs. Assembly Sequences
- working instructions
Sequences for assembling a block, consisting of a panel - lofting
stiffened by Iongitudinak and webs, can be classified as egg-
crate or weld Iongitudinar!s to panel first. The work se- - fabrication (marking, gas cutting, bending,
quences are different as shown in Figure 6-3. Thus, the line heating)
shrinkages caused by welding are sequenced differently. This - sub-block assembly
is important because restraints are different, the heat input - block assembly
for different welds varies and regions which have been shrunk
- shipwright work
- welding
that shipbuilders collect should be classified to match one or
more of the four assembly alternatives depicted in F@re 6-4.

6.9.4 Standards for Work Processes and Information Flow - check points
- dimensions to be checked
In order to establish effective standards, the role of each - checking methods
fabrication shop and assembly section must be carefully - tolerance limits
reviewed for its impact on production process flow. The in-
puts and outputs of each should be clearly defined and con- - checking prcxedures for jigs and machinery
sistent with a single A/C system. In other words, everything - feedback and remedial measures
on the right side of the variation merging equation must be
compatible in order to obtain the best productivity for the en-
tire hull construction process.

There cannot be dependence on just parochial knowledge. for flat block

Written work processes which regard each other are essential - plate arrangement (positioning, match mark)
for achieving specified accuracies and uniform flows of work
and information. Standard processes ako make it easier to - welding (misalignment, gap)
change jobs and are a great aid for training. When a proce-ss - panel marking (diagonal length, width, straightness)
standard is revised to incorporate an improvement, related - holes
others should be reviewed and/or revised as necessary.

The following guidance applies to standards which should for a curved block
be established: - supporting jig (normality, height)
- plate arrangement (jig position)
- datum line for joining
- block marking (four edges, diagonals)
- holes


1. Panel Assembly (E) 1. Panel Assembly (E)

2. Panel Marking 2. Panel Marking
3. Egg.crate Assembly (a,b,e) 3. Longitudinal to Panel Welding (A)
4. Eggcrate to Panel Welding (A) 4. Other-internals Welding (A,a,b,e)
5. Line Heating (B) if necessary 5. Line Heating (B) if necessary

FIGURE 6-3: When assembly sequences are different, tbe sequences of shrinkage and the amounts of shrinkage differ. Fitting problems will occur if dif-
ferent shrinkages are not antiapated. The parenthesized letters designate pertinent descriptions in Part 6.9.1.

for fitting I
- elimination of weldiug-bead rise where internals ems I
panel joints TEST STANDARD
- gas cutting (notch, roughness, check line)
- end of web position
- end of frame position MARKING
<1,000 mm s = 0.5 mm
HLINES *l,OLMJ mm s = 0.6 mm
- angle of internals relative to a panel
- collar-plate fitting a REFERENCE
- misalignment and gap where internals join each other LINE
a<l ,000 mm s = 0.4 mm
/ S= 0.4 mm
b aX,000 mm
- bead removal for rework
- bead removal to free temporary fitting a<1,000 mm s = 0.4 mm

line heating
- block interfaee edges
- specified temperatures
- specified locations
- fairing

FIGURE 6-5: Tests of measurement methods by one firm in-

dicated that even folding rulK do not cause significant variations.
However, each shipyard should perform smiliarr tests.

6.10 Measuring

Some variations are inevitable due to differences in:


Only At Egg-crate
Panel Edges I

One shipbuilding fm conducted tests of measurements

obtained with folding rules that are popular among shipbuild-
Proportionally ers. Of all devices folding rules were suspected of causing the
Throughout most measurement variation. The results, shown in Figure
6-5, indicate that even their use does not significantly contri-
bute to merged variation. However, each shipyard should
verify its own measuring eapabilities.

6.11 Photographs of A/C Practices

Figures 6-6 through 6-15 illustrate A/C ideas already

employed by shipbuilders to control accuracy and simultan-
eously enhance productivity.

FIGURE 6-4: There are two possible assembly sequences for each
of two methods for excess distribution.

A. B.

c. D.

FIGURE 6-11: Prudent A/C engineers assigned within a hdl construction department maintain good rapport with supervisors and their
workers. When variation merging equations identify a particular work process which needs to be improved such rapport leads to effective in-
novations; e.g., numerous simple jigs which significantly reduce ranges of variation.
A. Siple jigs support flat bars during the fitting process.
B. A jig which supports a flange during fitting to a web, is equipped with a screw for making fine adjustments.
C. Two relatively easy to make jigs align a small sub-block vertically and simultaneously fx the sub-block at the prescribed distance from
the panel edge.
D. Jigs are used to fit fongitudinals at prescribed angles during curved-block assembly. The jigs are designed so that they are suitable for
use on both forward or aft panel edges, and also on both port and starboard blocks.

c D

FIGURE 614: Typically A/C engineers establish standard procedures for:

A. Checking the face-plate position on a web.
B. Using a protractor and plumb bob form measuring the fitting angle of an internal member of a curved block.
C. Checking a bracket for fitting angle, match marks, etc.
D. Checking a finished edge for an erection butt-joint.


FIGURE 6-15: Specific A/C procedure apply to: A. Checking a measurement between panel and bracket edges m a sub-block.
B. Checking a measurement between the edges of a panel and internal structure in a block.
C. Checking alignment of a tank-top panel relative to a bottom-shell panel in a block.
D. Monitoring bottom alignment between keel laying and launching. The distance measured is that from the bottom shell to a reference
line marked on the vertical angle-iron which is fixed to the dock floor.



I. Identifying Vital Points

A. Basic
Vital points are necessary for achieving accuracy specified for an end product. Thus, identifying vital points starts
with the complete hull and proceeds, as any other planning activity, to address reverse production flow, i.e., erection,
block assembly, sub-block assembly and part fabrication. Also, because they impose different problems, each major
division of a ship body has its own vital-point explosion.
Vital points can be classified and sub-classified ax

1. At Erection Stage

a. Hold Zone
b. Curved Zone
c. Stern Zone

2. At Blcck Assembly Stage

a. Straight Block
b. Curved Block
c. Flat Panel Base
d. Curved Panel Base

3. At Part Fabrication

B. Detail Descriptions

1. Erection Stage
a. Hold Zone
Usually accuracy of the hold zone impacts most on the OVerall form of thhe hull because it contains the most
blocks. For vital-point matters, the hold zone can be subdivided into:

- Tank Top Zone

- Top Side Tank Zone
The tank top zone is the base of the hold and incorporates vital points for controlling

- Center line of the ship.

- Relativity between each double bottom block.
- Level of tank top.
See Attachment 1.
The top side tank zone fixes the actual width and add depth of the hull and contains vital points for con-

- Straightness of the base line.

- Width of the ship at main deck.
- Height of the ship at main deck.
- Level of main deck.
Details are shown in Attachment 2.
The vital points for setting each block on the ways is derived from the foregoing and noted for shipwright
guidance as shown in Attachment 3.

b. Curved Zone
Vital points in the curved zone are dependent on the hold zone because the block erection sequence usually
starts in the curved zone.

In order to set a curved block, fixing suitable points is necesary. For example:

Point A: For setting the width.

Point B: For keeping straightness.
Point C: For setting the height, and checking the lower width.

Note 1: Loftsmen must prepare dimension L to locate A on the shell:

Note 2: To locate point C, Loftsmen must provide dimensions H and B.


Stern Zone
Accuracy of the stem zone influences a ships performance significantly. Accuracy of the shaft line involves:
Accuracy of center of stem tube,
- Centering.
- Height.

Relationship between center of stem tube and the shaft line projected to the win engine seat

Notice: Keeping this relationship precise is especially hard because of movement of the stem block during welding.
Thus, fixing vital points and maintaining their positions requires the greatest possible care.

Usually the relationship between shaft and rudder centers are fixed in one block during block assembly. However, it is
still difficult to align both of them with sufficient accuracy in a building berth. The sequence for welding the plate joints
located forward of the after peak-tank bulkhead is critical.

2. Block Assembly & Stage

a. Straight Block
Straight bIocks are located in the hold zone, there are several typical types defined by their Iocations. In
order to define their vital points two questions should be asked:

A sample of a tyical check sheet is in Attachment 4.

b. Curved Block
Flat-panel base, curved blocks are assembled on a platen in accordance with a sequence which is partly
dependent upon internal structure.

As shown the curved shell plates are set on block internals. Therefore, vital points are set to maintain vital
dimensions such as A and H. The shell plate edge alignment with internal structure is aIso vital. See Attach-
ment 5.
Curved-panel base, curved blocks are assembled on a pin jig. The procedure is to first join already formed
plates to create a curved panel, layout the internal arrangement, and thereafter to fit and weld internals.
Typical vital points and dimensions and an applicable checking procedure are described in Attachment 6.

3. Part Fabrication Stage

As establishing vital points in all of the many parts is impractical, parts which could cause consequential block
inaccuracies are first identified. These typically are parts for:
- bottom girders
- bottom side floors
- hopper side tank floors
- hold frames

Vital point details and check sheets are provided in Attachment 7.

Appendix A, Attachment 1



In order to check and maintain accuracy of the tank top zone during the erection stage, three methods are necessary:

tank-top length.


Center Line Check

When: Twice, once before fitting and once after welding.
Who: Worker and A/C engineer before fitting.
A/C engineer after welding.
Where: At the front of each block on tank top.
How: By transit (allowance max. 1/8).

Relativity Check

When: Every block before fitting and once after welding an entire hold length.
Who: Worker and A/C engineer before fitting and A/C engineer after welding.
Where: At the front edge of each block.
H o w : By transit (allowance max. 1/8 at each target).
Notice: If the relativity is larger than allowed and that amount is less than 1/4, defer correction until welding is com-
plete for a hold length.

Level Check
When: Every block before fitting and after welding.
Who: Worker and A/C engineer before fitting and A/C engineer after welding.
Where: At points A, B, C and D at forward frame of each block on tank top.
After welding, the level of the points at the bottom must be checked:

The data should be recorded and arranged in a sirnpIe styIe (picture, graph, chart, etc.). Further each record should
contain the date, time, and temperature when the check was made. Recommended methods for recording these checks

Appendix A, Attachment 2



In order to check and maintain accuracy of the top side tank zone, four methods are necasary

l Straightness of the base line


1. Straightness of the Base Line

When: Twice, once before welding and once after welding at each erection joint-
Who: Worker and WC engineer before welding.
A/C engineer after welding.
Where: At the base line (see the figure at the end of this Attachment).
Notice: The base line must be marked on slabs before ereetion.
How: By transit.

2. Width of the Ship at Main Deck

Where Twice, before and after welding.
Who: Worker and A/C engineer before welding.
A/C engineer after welding.
Where: At the base line of the front part of block (see the figure at the end of this Attachment).
How: By measuring.

3. Height of the Ship at Main Deck

When: Twice, before and after welding.
Who: Worker and A/C engineer before welding.
A/C engineer after welding.
Where: At the point supported by the pillar (see the figure at the end of this Attachment).
How: By measuring.

4. Level of Main Deck

When: Twice, before and after welding.
Who: Worker and A/C engineer before welding.
A/C engineer after welding.
Where: At least 6 points as follows:

Notice: Points a & b at forward end.
Points c & d at aft end.
Points e & f at forward part of preceding block.
How: By transit.

Appendix A, Attachment 3


Appendix A, Attachment 4

Appendix A Attachment 5


Ship No. Block No. Shop Condition

232 (233)-1/2 Lower Engine Flat Base

Dim S Allow. Actual

Item in Chg. Notice
Tol . Dimension
Lower Eng. AC a-i 9 points
G Worker Keep horizontal plane
Flat Level

A Width Plumb at every frame

Vertical AC Plumb at every frame
H Height Worker Check the vertical

Each space at
F Space frame web

D Length

Notice 1) Keep the level and fix the flat panel.

2) Need support and strong back.


Ship N O . Block No. shop Condition

232(233)-2/2 Final Assembly

Notice After fitting and after welding.

Appendix A, Attachment 6
Curved Panel Base Blocks

1. A.C. Data Diagram

F I G . 1 FIG. 2

1. A. C. Data Diagram
It is generally difficult to check deformation of the curved unit shape. However, from the point of view of accuracy con-
trol it is necessary to check deformation of the curved unit shape during assembly work.

Then, the deformation checking data of the curved block shouId be prepared by the mold loft before they begin the
assembly work.
Calculate the maximum curvature depths at the aft butt, ford butt, upper erection seam, and the lower erection seam.
Join AD, BC, AB and CD as shown in Fig. 1.

Calculate the upper waterline sections depth and the lower waterline sections depth at the middle frame. And also
calculate the aft frame sections depth and ford frame sections depth at middle waterline.

Using the results of the above calculation, draw the checking data diagmm as shown in Fig. 2.

2. A.C. Checking Procedure



Performing basic statistical analyses requires understanding of three statistical concepts, mean, standard deviation and
the normal distribution curve. Consider the process of marking and cutting flat bars of identical nominal length. Each piece
has a measurable difference in length due to the inherent limitations of marking and cutting.

These two values, the mean and standard deviation, are for a random sample of size n. The random sample is taken from
the population of all flat bars produced by a specific process. The population can be considered as infnite in size, with the
random sample a finite subset. This sampling procedure can be repeated with a different batch of flat bars, measuring their
lengths and calculating a new mean and standard deviation. Generally, the means and standard deviations of the two ran-
dom sample will not be identical In theory, an infinite number of random samples of size n could be taken and their means
and standard deviations calculated. The laws of statistics state that the mean of all those means will be identical to the mean
of the entire population of the flat bars, i.e., all flat bars ever made by a specific, unchanged work process.

Raw data must be grouped to facilitate handling and analysis. Grouping data avoids the need for establishing precision
limits and has other advantages. Data grouping is done by measuring the length of each piece in the sample, arranging the
data into length classes, and determining the number of flat bars belonging to each class. The result is tabulated on a fre-
quency distribution table and is graphically represented by a frequency diagram or histogram.
The frequency distribution represents the number of occurrences of flat bars in each length class. Given a perfectly con-
trolled process, the frequency distribution will be a normal distribution. Where not perfectly controlled the frequency
distribution for a sampIe of measurements can be used to approximate the normal distribution for the process. The fol-
lowing table, histogram, and frequency distribution are examples associated with measuring the length of 100 flat bars:


Length Classes Midpoints Frequency (f)

(inches) (x) (number of pieces)

104.125-104.375 104.250 5
104.500-104.750 104.625
104.875-105.125 105.000 42
105.250-105.500 105.375 27
105.625-105.875 105.750 8

Sample size: 100

The area enveloped by the curve represents the total number in the sample. Generally, a distribution curve obtained from
actual data is not perfectly bell shaped as is the case for a normal distribution. As explained in Attachment 1, there is a way
to best fit a normal distribution and determine the pertinent risk factor.

tative length of all flat bars in the sample. When data is grouped by frequency of occurrence the mean is defined as:

The second fundamental parameter is standard deviation which is a measure of the dispersion or scatter of the observed
values around the mean value. If all observed lengths of flat bars tend to concentrate near the mean, the standard deviation
is small. If the values tend to be distributed far from the mean, the standard deviation is large.

Standard deviation is defined as:

for the example when x = 105.056

s = 0.365

Random variations from a well controlled process follow the normal distribution which is a symmetrical, bell-shaped
curve defined by its mean and standard deviation. The area beneath the curve always represents 100% of the sample being
considered apportioned as follows:

Value of x

These values can be obtained for any value of x from tables incorporated in statistics texts.


The distributing of controlled processes can be shown to be a normal distribution by applying the goodness-of-fit test as
a test for normaIity. This involves calculating the chi-square statistic:

where k = number of ranges in measured frequency distribution

Oi = frequency of observations in each range

in the following table confirm that the variations in shipbuilding work processes do follow the normal distribution. The
level of significance, a = 0.05, is the risk factor. That is, there is only a 5% chance that the goodness-of-fit test will indicate
a normal distribution when one does not exist.




Assembly Procedure

1. Fit the face plate to the web for the LL block shifted by S1.
2. Fit the face plate to the web for the LU block shifted by S2.

3. Fit the web to the panel for the LL block at A1 from the panel edge, where A1 = the design dimension + 2 mm.
4. Fit the web to the panel for the LU block at A2 from panel edge.

Variation Merging Equation

Z1 = ( A1 + E 1 ) A2

Ex1 and EX2 are inaccuracies due to curved deformation on inclination from vertical during erection which effects accur-
acy of the web gap Z2 near the face-plate side. Since it is difficult to obtain measurements of certain dimensions at the
erection site, Ex1 and Ex2 were calculated from the measured value of Z2 Z3 was calculated using Z2.

Sample size Mean value Vairiance

Dimension n Remarks
48 + 4.8 1-17 Right angle degree of upper end of LL web [After
cutting with edge extended 3mm (5rnm - 2mm) ]
-0.3 1.00 Right angle degree of lower end of LL web
+ 0.7 1.56 Fitting position of face plate to Web (LL)
+ 1.5 1.48 Fitting position of face plate to Web (LU)
+ 1.8 2.32 Fitting position of web frame LL (L) to panel
+ 0.6 2.48 Fitting position of LU web frame to Panel
- 1.6 2.91 Accuracy of seam joint of LL x LU (dimension between
reference lines after welding)
E XI + + 5.2 4.92
Estimated Gap Estimated back-strip welding 2.5%
Z1 -0.4 7.71
d o . 9%
Z3 - -0.5 17.84

Concepts are addressed in Appendix B and E.

Assembly Procedure

2. Provide 3 mm excess and finish cut fwd end of deck panel.

3. Provide some margin at the fwd end of the tank-bottom panel to be cut after the block is set during ereetion.

4. When joining the tank-bottom block with the deck block, align them by the distance A3.

Variation Merging Equations

Z1 = (A + E1 + A 2 ) - ( A 2 + P l )
Z 2 = ( A1 + E1 + A3 A4 ) - ( A3 + P2 A2 )




Variation Merging Equation



Variation Merging Equation

Z2 is more advantageous than equation Z, because there are fewer opportunities to generate variations. However, there
would be no advantage if the variations of P1 P2 and P3 were Small compared to A1, A2 and El. This type of analysis is
used to quantitatively determine the best design details for given production capabtities.


The number of processes required for the erection butt-joint in bottom longitudinal shown below is one less than for that
illustrated in Figure 3-2 of the basic text. Added processes sometimes increase merged variation at the final process.
However, an added process which does not contribute significantly to merged variation can be advantageous. In Figure 3-2,
the added prccess would permit the transverse bulkhead to be set more accurately.

Page D-2 of this Appendix contains a sample from the Japanese Shipbuilding Quality Standard (Hull Part) -1979;
published by the Research Committee on Steel Shipbuilding, The Society of Naval Architects of Japan, 15-16 Toranomon,
l-Chome, Minato-ku, Tokyo, Japan. Standard ranges and tolerance limits are identified for each item.

Pages D-3 through D-6 show how such accuracy standards were further developed by a shipbuilding firm.
Pages D-7 through D-10 are additional examples of independent accuracy standards development. These also specify
Frequency of Measurement.
Distortion & Straightness (Creature)


Bottom Shell

i i

Analysis of Shrinkage in Double-bottom Floor Caused by Gas Cutting and by Welding and Line Heating

Abilities to predict shrinkage caused by high temperatures and provide compensatory shrinkage allowances are crucial
for minimizing rework during erection. The problems are acerbated by the many different relatively complicated parts and
subassembly shapes that characterize shipbuilding. A double-bottom (DB) floor sub-bIock is a good example.

As shown in Figure 1, more than one DB floor panel is usually cut from a single plate. Shrinkage is different for the panel
edges which are different. The tortuous cutting paths for all 1 edges cause higher heat inputs. Thus, for each of them
shrinkage is greater than for any of the 2 and 3 edges which are straight.

Generally, except for more applied research, shipbuilders have done as much as can be done by modifying part shapes,
changing cutting-path sequences, and minimizing heat input. As shrinkage persists, it is counteracted by competitive ship-
builders with statistical methods for determining excess allowamces.
Figure 2 shows how A/C engineers have organized for and required the collection of statistical data following gas cutting
before a part is released from a part-fabrication shop to a sub-block assembly section. The data are organized as separate
histograms for each edge and provision is made to incorporate calculated mean values and standard deviations accordingly.
An example of how they were calculated is shown in Figure 3.
Data is again collected and analyzed in the same manner following line heating to remove distortion caused by welding
during sub-block assembly. The heat introduced by these two processes causes additional shrinkage, see Figures 4 and 5.
Also, the measurements to obtain these data serve as a check before a sub-block is released for block assembly.

However, the data recorded during sub-block assembly is an indicator of total shrinkage due to gas cutting plus welding
and line heating. Thus, it is necessary to calculate the mean values and standard deviations of just that shrinkage caused by
welding and line heating during sub-assembly as shown in Figure 6.

ferent from each other by less than 0.333 mm. The same can be said for the bulkhead and center edges.

side shows associated probabilities for rework. These percentages indicate that nearly 70% of the edges of all such sub-
block assemblies will make good connections.

Tolerance limits establish whether rework is necessary and indicate what kind of rework is required:


z, z2

However, for part dimensions which are are beyond the upper tolerance limit, i.e., too large, rework by gas cutting
should be deferred because during sub-block assembly

With A/C data accumulated during normal operations, statistics provides a way to predict the effect of a specific excess
allowance. The prediction is expressed as the percentages of parts which will during sub-block assembly require:

Typical actions which result from such predictions include:







Xi fi xifi
-6 1 - 6 36
-5 0 0 0
-4 3 -12 48
- 3 5 -15 45
- 2 5 - l 0 20
- 1 8 - 8 8
0 2 0 0
+ 1 3 3 3
+ 2 0 0 0
+3 I 3 9
+4 1 4 16
TOTAL 29 -41 185





In this example, variation from the design dimension beyond 4mm requires back-strip welding and variation beyond
+ 2mm always requires gas-cutting. The possibly no rework range applies to variadons between O and + 2mm for which
the upper limit was established based upon the shipyards experience. The values of Fn used to obtain the percentages of
work for each region, are calculated using a normal distribution with the means and standard deviations obtained as shown
in Figures 1 through 6 for the region limits given above.





Accuracy control (A/c) is based on the variation of products manufactured in the same manner. Even for controll-
ed processes, i.e., where work circumstances do not change, some chance or random variation is normal. As variation
is expected, A/C is also concerned with detecting when a process is deviating from its controlled condition. In other
words, A/C engineers must be alert for variations which are not due to chance as they are indicators that something
or someone is changing how work is being performed.
A/C engineers employ two kinds of charts for control purposes. One is for measurements such as the lengths of flat
bars and the other addresses frequencies or counted data, e.g., the number in a sample of 100 that require rework.
Both charts employ central lines indicating the average performance expected of a process and upper and lower
control-limit lines. The Iimits are chosen so that values between them represent only normal, random variation.
Values beyond the upper and lower control limits indicate that a work process is out of control. B Y plotting values of
samples taken periodically, A/C engineers can also detect a drift toward loss of control.
(s). AS the population mean and standard deviation are generally unknown, they are estimated by first obtaining a

and then assumed to be the estimated population mean.

In order to estimate the standard deviation of the population it is necessary to calculate the standard deviation for
each sample. Because it is easier to obtain, range (R) is a preferred indicator. The range of variation of each sample
(Ri) is used to calculate the mean range:

from a Table of Control Chart Constants (See ASTM Manual on Quality Control of Materials, American Society for

A similar approach is used for the range or R chart for which:

Examples of control charts follow:




Additional copies of this report can be obtained from the
National Shipbuilding Research and Documentation Center:

Documentation Center
The University of Michigan
Transportation Research Institute
Marine Systems Division
2901 Baxter Road
Ann Arbor, MI 48109-2150

Phone: 734-763-2465
Fax: 734-763-4862